Estimating Incidence Rates in the Evaluation of Pooled Safety Data

Abstract

Tools developed for survival analysis, such as the estimation of incidence rates, have lately been used to evaluate adverse drug reactions; this approach has been favored by the regulatory authorities. Acceptance in particular of the concept of incidence rate by readers without statistical training is, however, limited because of it's confusion with the more intuitive concept of the conditional probability of the initial occurrence of an abnormality after baseline. It is shown that in many practical examples, the incidence rate and the conditional probability provide simple approximations of each other.

Another reason for limited acceptance is that some of the properties of the estimators available in survival analysis are hard to interpret in the context of pooled safety evaluation; this is demonstrated for the central hazard rate, a modification of it is proposed for the analysis of safety data pooled from clinical trials with different visit structures.